Abstract
Natural language processing (NLP) is the field that enables computers to understand and use human language. Textual entailment—a key NLP task— determines if a hypothesis can logically follow from a given premise. As we reviewed, the model designed and developed for other languages is not used for Afaan Oromoo textual entailment classification, as its semantics and syntax are different when compared with other languages. To address the gap, we proposed an Afaan Oromoo textual entailment classification model. We used Support Vector Machine (SVM) as a baseline to compare with three deep learning architectures: Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Bidirectional Long Short-Term Memory (BiLSTM) by comparing their performance to identify the most effective approach with fasttext and word2vec word embedding. We collected a dataset of 13,060 sentence pairs in Afaan Oromoo. The accuracy of SVM was 55.82% and the accuracy of CNN, LSTM, and BiLSTM was 72.8%, 75.57% and 80.47% respectively, with fasttext word embedding. Considering the limited resources available for Afaan Oromoo NLP, the result is encouraging. As a starting point, this study offers a basis for additional investigation and advancement in this field and contributes to the development of Afaan Oromoo's Natural Language Processing capabilities.